Source access control
---------------------
`active_docs` is client-supplied and reached the retriever unchecked, and the
retriever queries `WHERE source_id = <id>` with no owner predicate — so any
caller could pass any source id to /stream or /api/answer and have another
tenant's documents quoted back, while /api/sources/<id>/search correctly
refused the same id. Gate it through `can_access`, the helper the guarded
endpoints already use, and filter `self.source` down to the authorized set.
Fails closed: no principal, or a check that errors, drops the source.
Three sibling paths had the same gap:
- workflow agent nodes: `AgentNodeConfig.sources` is written verbatim from
client JSON at save time and nothing validated it, so a node could name any
tenant's source. Gate against the workflow owner, so shared workflows keep
reading their owner's sources like shared agents do.
- /api/share: `_resolve_source_pg_id` resolved any id with no ownership
predicate and baked it into the agent the share creates; /api/search then
searched it. Authorize before attaching.
- search_service: re-resolve the ids stored on an agent row instead of
trusting them, so a row written by any future path with the same gap cannot
be read back.
Team grantees previously lost their source's retrieval config: the post-check
read was still owner-scoped, so it missed and fell back to defaults (an
`agentic_tool` source was bulk-prefetched for every grantee). Read unscoped
after `can_access` passes.
Retrieval
---------
`PGVectorStore._ensure_table_exists` created an IVFFlat index on the empty
table it had just created. IVFFlat computes centroids at build time, so those
centroids were random, and combined with the `source_id` post-filter a source
with hundreds of embedded chunks returned zero rows — retrieval reported no
documents, the model answered from memory, and nothing was logged. Stop
creating the index (exact search is correct and fast well past the sizes most
deployments reach); raise `ivfflat.probes` to sqrt(lists) where an index still
exists; and re-run a short indexed search exactly, since post-filtering means
no index setting can guarantee a full result. `graphrag` had the same
empty-table index with no fallback at all.
Also: bound `chunks` to 0-500 on both the request and agent paths (0 still
means "skip retrieval"), let a source's configured `retrieval.chunks` outrank
the request body, and cap ClassicRAG's per-source floor at
max(top_k, n_sources) so attaching sources cannot inflate the result set.
Silent failures
---------------
An empty retrieval was invisible to both the model and the client: the `source`
event was suppressed when the list was empty, so "searched and found nothing"
looked identical to "no source attached", and the prompt said nothing at all.
Emit the event always, and tell the model when a search ran and returned
nothing. A file that parses to nothing now fails ingest with a message naming
the cause instead of storing an embedding of the empty string. `score_threshold`
returns warnings when the active store or retriever cannot honour it.
Prompt structure
----------------
Retrieved documents move from the system prompt into the user turn, with the
injection guard restated next to them: they change every turn (defeating prefix
caching), they are third-party text that should not carry system authority, and
routing them through the query budget makes them truncatable rather than
silently crowding it out. Documents are shed lowest-ranked-first before the
question is touched.
The six chat presets (3 tones x 2 retrieval modes) differed only in their
Answering section; they are now composed from single-source fragments at load
time, not through Jinja inheritance, which would have opened a file-read
surface in the template sandbox and broken the tool-prefetch parser. Per-tool
guidance moves out of the prompt into tool schemas, so it travels with the tool
and cannot render when the tool is absent. A plain-text custom prompt is staged
as a persona value inside the skeleton instead of replacing it wholesale — it
used to silently lose the injection guard, platform block, memory and
attachments, and its braces are now inert.
Other fixes
-----------
- agents/base: an oversized system prompt drove the query budget negative and
dispatched a full-price request with an empty question; raise instead.
- llm/anthropic: migrate off the retired Text Completions API. It flattened
history to first+last message and ignored tools entirely. Adds the missing
Anthropic handler, without which every tool call was silently dropped.
- sources/upload: `sitemap` had no branch, so every sitemap ingest died on a
TypeError; `validate_url` now rejects a falsy URL cleanly.
- workflow nodes: retrieved documents never reached the node agent, so a
classic node with a source and an ordinary prompt answered "I have no
documents" while the run reported completed.
- parser/bulk: copy the metadata dict, or every chunk reports the last chunk's
token_count.
- crawler_loader: carry the page title, or citations render the whole chunk
body as the label.
WorkflowEngine reports node failures by *yielding* `{"type": "error"}`
rather than raising, so complete_stream's generator returns normally and
the except handler never runs. The turn was finalized `status="complete"`
with an empty response.
Live, the client renders an error bubble with a Retry button. On reload
it does not: mapServerQueryToClient only surfaces `metadata.error` for
`failed` rows, so history showed a blank message with no error and no way
to retry. A user hitting this re-sends the same prompt into new
conversations, which is exactly what the 2026-08-01 report shows — nine
blank first messages in seven hours.
Tracks a `stream_error` flag alongside the existing `paused` machinery
and finalizes `failed` when the turn produced no answer, recording the
user-facing message in `metadata.error`. An error arriving *after* output
keeps `complete` so partial text is not discarded; structured answers
count as output too, since they live in `structured_chunks` rather than
`response_full`. The flag is recorded before the pause branches so those
paths cannot lose it.
save_conversation grew a `status` parameter (default `complete`) for the
non-WAL branch, which took no status and so landed on the column default
— the same blank-complete row on a path the WAL fix did not cover.
Title generation now also runs for failed turns: _maybe_generate_title
only regenerates while the name is still the question-prefix fallback, so
skipping it would strand a conversation whose first turn failed with the
raw prompt as its name forever.
logging.py counts a yielded error toward `activity_finished.status`.
These failures previously logged `status=ok` with `answer_length=0`,
which is why user-visible blank answers never appeared in error metrics.
Added button to rotate api key. This is required for better UX, now user has to remake agent if key is lost.
A migration is needed to keep track of proper usage + to not leave any orphan entries.
Token accounting:
- Drain each tool round's provider stream to exhaustion before running
tools and recursing, so the usage decorator persists exactly one
token_usage row per LLM call, at call end. Previously every round's
generator was abandoned mid-iteration and flushed together at request
teardown, writing N near-identical rows stamped with the final
round's provider counts (duplicate billing).
- Consume the Chat Completions include_usage terminal chunk (it arrives
after finish_reason and was never read) so streamed calls record
provider-exact token counts instead of tiktoken estimates.
- Claim provider-reported usage once per call (_last_usage_claimed) so
a late-finalized generator can never adopt another call's counts.
Oversized-context guards:
- Enforce Responses API function_call/function_call_output pairing in
the input builder (drop unpaired items; bypassed for store-mode
previous_response_id chaining where calls are matched server-side).
- Hard pre-send context gate: shrink oversized tool results and refuse
payloads that cannot fit the model's window before dispatch, so a
hopeless request is never sent or billed.
- Cap a single tool result entering the LLM context
(TOOL_RESULT_MAX_TOKENS, default 20000); the tool journal and
persistence keep the full result. Applied on the resume/continuation
path too.
- Skip the fallback attempt when the payload cannot fit the fallback
model's context window (10% estimation slack).
- Compression: never save a compression point that does not reduce
tokens; bound oversized verbatim fields kept after a compression
point (COMPRESSION_RECENT_FIELD_MAX_TOKENS, default 8000).
Robustness fixes from review:
- Google parallel function calls: complete index-less ToolCalls are no
longer merged into one another (dict arguments raised TypeError on
+=; second call could execute with the first call's arguments).
- Trailing-frame failures after a delivered answer no longer error the
stream or restream the whole answer from the fallback
(_stream_reached_finish).
- In-memory compression falls back to minimal pruning when the summary
is not smaller than the original.
- keep<=0 guard in the middle-truncation helpers (a tiny cap returned
marker + full text).
Frontend: tooltip on the Analytics tokens stat card explaining that
agent tool loops re-send conversation context on every step.
Wrap a storage read in a context manager (close the file handle), add a comment
on the expected SandboxCapacityError in the churn test, and unify the
artifacts-routes test on a single import style. Also annotate the intentional
per-session sandbox workspace path with # nosec B108 (controlled dir inside the
runner container, not an insecure shared temp file).
Replace the sandbox Docling extractor with read_document, backed by the in-process
backend parser (the same one ingestion uses) and offloaded to a dedicated
'parsing' Celery queue so it can run on GPU-capable workers with predictable RAM.
The tool resolves the input ref under the run-scoped gate, enqueues the parse,
and awaits it with a timeout (degrading to an error rather than hanging); the
worker independently re-resolves the artifact through the same gate and never
trusts a raw path. Untrusted files get the upload path's safeguards (extension
whitelist, size cap, sanitized temp file, cleanup). Options: output
(markdown/text/structured/chunks), ocr, pages, engine, max_chars, include_tables,
persist, json_schema. The workflow native-file 'extract' fallback now uses the
same worker path, so document parsing no longer needs the sandbox and works on
every backend.
Also fixes the branch's periodic-task test (the sandbox reaper made it 12) and
points the dev and e2e Celery workers at the parsing queue.
The classic prompt renderer now passes artifact_parent={conversation_id} so a
normal agent's prompt can resolve a prior-turn artifact with
{{ artifacts.artifact(id) }}, scoped to its own conversation (parent-derived
authz; a missing conversation_id safely yields an empty lookup). This makes the
artifacts variable surfaced in the agent builder functional, and is the parent
wiring an agent-as-tool / subagent feature would also rely on.
Let workflow runs consume and produce documents end to end: bridge uploaded
attachments into run-scoped artifacts so nodes receive the input documents
(with a per-run cap and server-computed size/sha256, and the run row pre-created
so produced artifacts are authorized during the run); emit the run id to the
client and add a builder panel that lists, previews, and downloads a run's
artifacts; and allow attaching documents to a Preview run via the existing
upload flow.
Also fixes issues a compliance workflow surfaced: attachment ownership now keys
on the raw identity instead of a sanitized one (the sanitized form could not be
read back and could collide across users); workflow code nodes read prior state
from a state.json data file instead of templating it into the program, so
untrusted document content can never be interpolated into executed code;
structured node output wrapped in code fences is recovered; and the live
speech-to-text ownership check compares the raw identity.
Alex